Implications of automated ontology identification — LessWrong